Deep Learning excercise on Severstal dataset from Kaggle competition https://www.kaggle.com/c/severstal-steel-defect-detection
Learning objectives:
- perform data exploration
- build deep learning model using U-Net for Image Segmentation
- learn/use keras and tensorflow
- using Colab for running Notebook with GPU
- Downloading data in Colab using Kaggle API https://towardsdatascience.com/downloading-datasets-into-google-drive-via-google-colab-bcb1b30b0166
- how to build and visualize masks https://www.kaggle.com/titericz/building-and-visualizing-masks
- Building Keras data generator https://stanford.edu/~shervine/blog/keras-how-to-generate-data-on-the-fly
- Model architecture https://www.kaggle.com/xhlulu/severstal-simple-keras-u-net-boilerplate
Use Google Colab
Make sure to use GPU as Hardware Accelerator in Edit -> Notebook settings
Results validation